Statistical Evaluation of Outdoor Field Hockey Penalty Corners

Penalty corners stand out as pivotal goal-scoring opportunities in field hockey, crucial to a team’s triumph. This study harnesses data from women’s collegiate field hockey games to formulate a statistical model predicting the likelihood of scoring a penalty corner, contingent on the strategies deployed. Various machine learning algorithms are compared to ascertain the most predictive model and to dissect the paramount factors influencing penalty corners. The XGBoost model emerges superior, boasting an area under the curve (AUC) score of 0.667 on out-of-sample observations. With other predictors held constant, the model reveals that drag flicks, sweep shots, and deflections are positively associated with goal occurrences, while, intriguingly, direct shots—despite their prevalence—are negatively associated with scoring probability. KEYWORDS: College Sports; K-Nearest Neighbor; Lasso; Quantitative Analysis; Random Forest; Sporting Strategy; Sports Analytics; XGBoost